Currently, standardization of encoding system called High Efficiency Video Coding (HEVC) is promoted by Joint Collaboration Team-Video Coding (JCTVC), a joint standardization organization of ITU-T and ISO/IEC, aiming at further improvement of H. 264/AVC in encoding efficiency. Concerning HEVC standards, a committee draft as the initial draft version is issued in February, 2012 (e.g., see Non Patent Literature 1).
According to HEVC standards, a picture can be split into units of tiles or slices for encoding. In decoding an encoded stream split into these units and encoded, no correlation exists between the split units in the process of creating information about Context-based Adaptive Binary Arithmetic Coding (CABAC), intra prediction modes, quantization values, and the like.
However, according to inter prediction, no restriction is set to motion vectors. In this case, an encoded image of a different tile at a different time can be used as a reference image. Accordingly, independent encoding and decoding in the time direction for each tile is not allowed.
More specifically, as shown in FIG. 1, for example, each of a frame #t having a Picture Order Count (POC) of t and a frame #t−1 having a POC of t−1 is split into four tiles and inter-predicted, all the encoded images within the four tiles of the frame #t−1 can be determined as possible reference images for a CU (Coding Unit) of the frame #t.
Accordingly, there is a case when a decoded image 12 within a tile #2 having a specific ID (hereinafter referred to as a tile ID) of 2 and contained in the frame #t−1 is determined as a reference image for a CU 11 of a tile #1 having a tile ID of 1 and contained in the frame #t, for example. In other words, there is a case when a vector which has an initial point of a CU 11 and a terminal point of an area 12A of the frame #t corresponding to the decoded image 12 is detected as a motion vector 13. In this case, reference to the decoded image 12 of the tile #2 different from the tile #1 containing the CU 11 is needed; therefore, independent encoding and decoding in the time direction for each tile is not allowed.
Accordingly, a decoding device needs to have a common decoding Decoded Picture Buffer (DPB) which retains decoded images for all tiles.
FIG. 2 is a block diagram showing a constitution example of a decoding device of this type.
A decoding device 30 in FIG. 2 is constituted by decoding units 31-1 through 31-N, DPB 32-1 through 32-N, and a common DPB 33.
An encoded stream containing split N tiles (N is an arbitrary positive number) and encoded for each unit of tiles is inputted to the decoding device 30. Encoded data of each tile is supplied to the corresponding decoding units 31-1 through 31-N.
Each of the decoding units 31-1 through 31-N decodes the encoded data of the corresponding tile by using the corresponding image of decoded images stored in the common DPB 33 for all tiles contained in the corresponding frame as a reference image.
More specifically, the decoding unit 31-1 decodes encoded data of a tile #1 having a tile ID of 1 using a reference image, and supplies a decoded image of the tile #1 obtained as a result of the decoding to the DPB 32-1. Similarly, the decoding units 31-2 through 31-N decode data of a tile #2 having a tile ID of 2, a tile #3 having a tile ID of 3, and up to a tile #N having a tile ID of N using reference images, respectively. Then, the decoding units 31-2 through 31-N supply the decoded images of the tile #2, tile #3, and up to tile #N obtained by decoding to the DPB 32-2, DPB 32-3, and up to DPB 32-N, respectively.
The DPB 32-1 through 32-N store the decoded images supplied by the corresponding decoding units 31-1 through 31-N. The DPB 32-1 through 32-N supply the stored decoded images to the common DPB 33 and allow these decoded images to be stored in the common DPB 33.
The common DPB 33 stores the decoded images of the tile #1 through the tile #N at the same time supplied by the DPB 32-1 through 32-N as decoded images of one frame. The common DPB 33 outputs the stored decoded images for each unit of frames as decoded results.
In addition, though not shown in the figures, a common DPB needs to be provided on an encoding device for inter prediction similarly to the DPB on the decoding device 30.